DESCRIPTION (From applicant s abstract): This continuing project concerns further work on the problem of measurement error in general regression problems. The developed methodology will be applied to a number of data sets that the investigators have in their possession. The research to be performed falls into three broad categories. 1) Analysis of dietary intake data. Calibration studies that relate FFQs to usual intake will be investigated. Missing data and optimal semiparametric methods will be discussed. Estimating the distribution of usual intake will be studied. A new method, called nonparametric calibration, will be developed for relating dietary instruments to usual intake as an unknown function of covariates such as body mass index. 2) Functional and structural nonlinear measurement error modeling. New general methodology will be developed for nonlinear measurement error modeling when no assumptions are made about the underlying distribution of the error-prone predictor. 3) Semiparametric and dimension reduction methods.: Motivated by problems in measurement error and missing data, the investigators will develop a general theory of semiparametric plug-in estimation, in which a semiparametric function is estimated and substituted into an unbiased estimating equation. Within the Generalized Partially Linear Single-Index Model (GPLSIM) family, results will be extended from quasi-likelihood models to general regression problems. An important application of this extension will be to nonparametric calibration in dietary intake studies.